diff mbox series

xfrm: Remove inner/outer modes from input path

Message ID ZAr3rc+QvKs50xkm@gondor.apana.org.au (mailing list archive)
State Awaiting Upstream
Delegated to: Netdev Maintainers
Headers show
Series xfrm: Remove inner/outer modes from input path | expand

Checks

Context Check Description
netdev/series_format warning Single patches do not need cover letters; Target tree name not specified in the subject
netdev/tree_selection success Guessed tree name to be net-next
netdev/fixes_present success Fixes tag not required for -next series
netdev/header_inline success No static functions without inline keyword in header files
netdev/build_32bit success Errors and warnings before: 23 this patch: 23
netdev/cc_maintainers warning 4 maintainers not CCed: kuba@kernel.org edumazet@google.com pabeni@redhat.com davem@davemloft.net
netdev/build_clang success Errors and warnings before: 18 this patch: 18
netdev/verify_signedoff success Signed-off-by tag matches author and committer
netdev/deprecated_api success None detected
netdev/check_selftest success No net selftest shell script
netdev/verify_fixes success No Fixes tag
netdev/build_allmodconfig_warn success Errors and warnings before: 23 this patch: 23
netdev/checkpatch success total: 0 errors, 0 warnings, 0 checks, 157 lines checked
netdev/kdoc success Errors and warnings before: 0 this patch: 0
netdev/source_inline success Was 0 now: 0

Commit Message

Herbert Xu March 10, 2023, 9:26 a.m. UTC 
The inner/outer modes were added to abstract out common code that
were once duplicated between IPv4 and IPv6.  As time went on the
abstractions have been removed and we are now left with empty
shells that only contain duplicate information.  These can be
removed one-by-one as the same information is already present
elsewhere in the xfrm_state object.

Removing them from the input path actually allows certain valid
combinations that are currently disallowed.  In particular, when
a transport mode SA sits beneath a tunnel mode SA that changes
address families, at present the transport mode SA cannot have
AF_UNSPEC as its selector because it will be erroneously be treated
as inter-family itself even though it simply sits beneath one.

This is a serious problem because you can't set the selector to
non-AF_UNSPEC either as that will cause the selector match to
fail as we always match selectors to the inner-most traffic.

Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>

Comments

Steffen Klassert March 15, 2023, 11:05 a.m. UTC | #1 
On Fri, Mar 10, 2023 at 05:26:05PM +0800, Herbert Xu wrote:
> The inner/outer modes were added to abstract out common code that
> were once duplicated between IPv4 and IPv6.  As time went on the
> abstractions have been removed and we are now left with empty
> shells that only contain duplicate information.  These can be
> removed one-by-one as the same information is already present
> elsewhere in the xfrm_state object.
> 
> Removing them from the input path actually allows certain valid
> combinations that are currently disallowed.  In particular, when
> a transport mode SA sits beneath a tunnel mode SA that changes
> address families, at present the transport mode SA cannot have
> AF_UNSPEC as its selector because it will be erroneously be treated
> as inter-family itself even though it simply sits beneath one.
> 
> This is a serious problem because you can't set the selector to
> non-AF_UNSPEC either as that will cause the selector match to
> fail as we always match selectors to the inner-most traffic.
> 
> Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>

Applied to ipsec-next, thanks a lot Herbert!
Steffen Klassert June 6, 2023, 10:45 a.m. UTC | #2 
Hi Herbert,

On Fri, Mar 10, 2023 at 05:26:05PM +0800, Herbert Xu wrote:
...
> @@ -369,17 +366,12 @@ static int xfrm_prepare_input(struct xfrm_state *x, struct sk_buff *skb)
>  		return -EAFNOSUPPORT;
>  	}
>  
> -	if (x->sel.family == AF_UNSPEC) {
> -		inner_mode = xfrm_ip2inner_mode(x, XFRM_MODE_SKB_CB(skb)->protocol);
> -		if (!inner_mode)
> -			return -EAFNOSUPPORT;
> -	}
> -
> -	switch (inner_mode->family) {
> -	case AF_INET:
> +	switch (XFRM_MODE_SKB_CB(skb)->protocol) {
> +	case IPPROTO_IPIP:
> +	case IPPROTO_BEETPH:

the assumption that the L4 protocol on BEET mode can be
just IPIP or BEETPH seems not to be correct. One of
our testcaces hit the second WARN_ON_ONCE() in
xfrm_prepare_input. In that case the L4 protocol
is UDP. Looks like we need some other way to
dertermine the inner protocol family.
diff mbox series

Patch

diff --git a/net/xfrm/xfrm_input.c b/net/xfrm/xfrm_input.c
index 436d29640ac2..39fb91ff23d9 100644
--- a/net/xfrm/xfrm_input.c
+++ b/net/xfrm/xfrm_input.c
@@ -231,9 +231,6 @@  static int xfrm4_remove_tunnel_encap(struct xfrm_state *x, struct sk_buff *skb)
 {
 	int err = -EINVAL;
 
-	if (XFRM_MODE_SKB_CB(skb)->protocol != IPPROTO_IPIP)
-		goto out;
-
 	if (!pskb_may_pull(skb, sizeof(struct iphdr)))
 		goto out;
 
@@ -269,8 +266,6 @@  static int xfrm6_remove_tunnel_encap(struct xfrm_state *x, struct sk_buff *skb)
 {
 	int err = -EINVAL;
 
-	if (XFRM_MODE_SKB_CB(skb)->protocol != IPPROTO_IPV6)
-		goto out;
 	if (!pskb_may_pull(skb, sizeof(struct ipv6hdr)))
 		goto out;
 
@@ -331,22 +326,26 @@  static int xfrm6_remove_beet_encap(struct xfrm_state *x, struct sk_buff *skb)
  */
 static int
 xfrm_inner_mode_encap_remove(struct xfrm_state *x,
-			     const struct xfrm_mode *inner_mode,
 			     struct sk_buff *skb)
 {
-	switch (inner_mode->encap) {
+	switch (x->props.mode) {
 	case XFRM_MODE_BEET:
-		if (inner_mode->family == AF_INET)
+		switch (XFRM_MODE_SKB_CB(skb)->protocol) {
+		case IPPROTO_IPIP:
+		case IPPROTO_BEETPH:
 			return xfrm4_remove_beet_encap(x, skb);
-		if (inner_mode->family == AF_INET6)
+		case IPPROTO_IPV6:
 			return xfrm6_remove_beet_encap(x, skb);
+		}
 		break;
 	case XFRM_MODE_TUNNEL:
-		if (inner_mode->family == AF_INET)
+		switch (XFRM_MODE_SKB_CB(skb)->protocol) {
+		case IPPROTO_IPIP:
 			return xfrm4_remove_tunnel_encap(x, skb);
-		if (inner_mode->family == AF_INET6)
+		case IPPROTO_IPV6:
 			return xfrm6_remove_tunnel_encap(x, skb);
 		break;
+		}
 	}
 
 	WARN_ON_ONCE(1);
@@ -355,9 +354,7 @@  xfrm_inner_mode_encap_remove(struct xfrm_state *x,
 
 static int xfrm_prepare_input(struct xfrm_state *x, struct sk_buff *skb)
 {
-	const struct xfrm_mode *inner_mode = &x->inner_mode;
-
-	switch (x->outer_mode.family) {
+	switch (x->props.family) {
 	case AF_INET:
 		xfrm4_extract_header(skb);
 		break;
@@ -369,17 +366,12 @@  static int xfrm_prepare_input(struct xfrm_state *x, struct sk_buff *skb)
 		return -EAFNOSUPPORT;
 	}
 
-	if (x->sel.family == AF_UNSPEC) {
-		inner_mode = xfrm_ip2inner_mode(x, XFRM_MODE_SKB_CB(skb)->protocol);
-		if (!inner_mode)
-			return -EAFNOSUPPORT;
-	}
-
-	switch (inner_mode->family) {
-	case AF_INET:
+	switch (XFRM_MODE_SKB_CB(skb)->protocol) {
+	case IPPROTO_IPIP:
+	case IPPROTO_BEETPH:
 		skb->protocol = htons(ETH_P_IP);
 		break;
-	case AF_INET6:
+	case IPPROTO_IPV6:
 		skb->protocol = htons(ETH_P_IPV6);
 		break;
 	default:
@@ -387,7 +379,7 @@  static int xfrm_prepare_input(struct xfrm_state *x, struct sk_buff *skb)
 		break;
 	}
 
-	return xfrm_inner_mode_encap_remove(x, inner_mode, skb);
+	return xfrm_inner_mode_encap_remove(x, skb);
 }
 
 /* Remove encapsulation header.
@@ -433,17 +425,16 @@  static int xfrm6_transport_input(struct xfrm_state *x, struct sk_buff *skb)
 }
 
 static int xfrm_inner_mode_input(struct xfrm_state *x,
-				 const struct xfrm_mode *inner_mode,
 				 struct sk_buff *skb)
 {
-	switch (inner_mode->encap) {
+	switch (x->props.mode) {
 	case XFRM_MODE_BEET:
 	case XFRM_MODE_TUNNEL:
 		return xfrm_prepare_input(x, skb);
 	case XFRM_MODE_TRANSPORT:
-		if (inner_mode->family == AF_INET)
+		if (x->props.family == AF_INET)
 			return xfrm4_transport_input(x, skb);
-		if (inner_mode->family == AF_INET6)
+		if (x->props.family == AF_INET6)
 			return xfrm6_transport_input(x, skb);
 		break;
 	case XFRM_MODE_ROUTEOPTIMIZATION:
@@ -461,7 +452,6 @@  int xfrm_input(struct sk_buff *skb, int nexthdr, __be32 spi, int encap_type)
 {
 	const struct xfrm_state_afinfo *afinfo;
 	struct net *net = dev_net(skb->dev);
-	const struct xfrm_mode *inner_mode;
 	int err;
 	__be32 seq;
 	__be32 seq_hi;
@@ -491,7 +481,7 @@  int xfrm_input(struct sk_buff *skb, int nexthdr, __be32 spi, int encap_type)
 			goto drop;
 		}
 
-		family = x->outer_mode.family;
+		family = x->props.family;
 
 		/* An encap_type of -1 indicates async resumption. */
 		if (encap_type == -1) {
@@ -676,17 +666,7 @@  int xfrm_input(struct sk_buff *skb, int nexthdr, __be32 spi, int encap_type)
 
 		XFRM_MODE_SKB_CB(skb)->protocol = nexthdr;
 
-		inner_mode = &x->inner_mode;
-
-		if (x->sel.family == AF_UNSPEC) {
-			inner_mode = xfrm_ip2inner_mode(x, XFRM_MODE_SKB_CB(skb)->protocol);
-			if (inner_mode == NULL) {
-				XFRM_INC_STATS(net, LINUX_MIB_XFRMINSTATEMODEERROR);
-				goto drop;
-			}
-		}
-
-		if (xfrm_inner_mode_input(x, inner_mode, skb)) {
+		if (xfrm_inner_mode_input(x, skb)) {
 			XFRM_INC_STATS(net, LINUX_MIB_XFRMINSTATEMODEERROR);
 			goto drop;
 		}
@@ -701,7 +681,7 @@  int xfrm_input(struct sk_buff *skb, int nexthdr, __be32 spi, int encap_type)
 		 * transport mode so the outer address is identical.
 		 */
 		daddr = &x->id.daddr;
-		family = x->outer_mode.family;
+		family = x->props.family;
 
 		err = xfrm_parse_spi(skb, nexthdr, &spi, &seq);
 		if (err < 0) {
@@ -732,7 +712,7 @@  int xfrm_input(struct sk_buff *skb, int nexthdr, __be32 spi, int encap_type)
 
 		err = -EAFNOSUPPORT;
 		rcu_read_lock();
-		afinfo = xfrm_state_afinfo_get_rcu(x->inner_mode.family);
+		afinfo = xfrm_state_afinfo_get_rcu(x->props.family);
 		if (likely(afinfo))
 			err = afinfo->transport_finish(skb, xfrm_gro || async);
 		rcu_read_unlock();